hard to believe

[baltimorened]

According to a Harvard Harris poll, Democrats and republicans are just about even in a generic comparison. This appears to be a waaay outlier to what other polls are showing.

here's more from the report:

Fifty percent of voters say their minds are already made up about who to vote for and two thirds say they expect to vote a straight-party ticket. Fifty-nine percent of those supporting Democrats say they have definitely made up their minds, and 58 percent of Republicans say the same. Just 33 percent of independent-leaning voters say their minds are made up, with 26 percent saying they are leaning in one direction and 24 percent said they have not decided.

 

[MTTiger19]

According to a Harvard Harris poll, Democrats and republicans are just about even in a generic comparison. This appears to be a waaay outlier to what other polls are showing.

here's more from the report:

Fifty percent of voters say their minds are already made up about who to vote for and two thirds say they expect to vote a straight-party ticket. Fifty-nine percent of those supporting Democrats say they have definitely made up their minds, and 58 percent of Republicans say the same. Just 33 percent of independent-leaning voters say their minds are made up, with 26 percent saying they are leaning in one direction and 24 percent said they have not decided.

Division and polarization are at ATH baby. Breaking point soon?

 

[Scrubby]

Modern polling is almost completely worthless. If you factor in the same democrat lean that polls have had for the last few election cycles then republicans are currently on track to keep both the house and the senate.

 

[Gibbs6]

Modern polling is almost completely worthless. If you factor in the same democrat lean that polls have had for the last few election cycles then republicans are currently on track to keep both the house and the senate.

I will never understand these worthless pre election polls. I’ll get sent texts to take these polls every now and then. If for some reason I’m bored enough to fill one out, I never respond truthfully to the poll. I find it more enjoyable to **** with the results as they are meaningless.

 

[FLaw47]

According to a Harvard Harris poll, Democrats and republicans are just about even in a generic comparison. This appears to be a waaay outlier to what other polls are showing.

here's more from the report:

Fifty percent of voters say their minds are already made up about who to vote for and two thirds say they expect to vote a straight-party ticket. Fifty-nine percent of those supporting Democrats say they have definitely made up their minds, and 58 percent of Republicans say the same. Just 33 percent of independent-leaning voters say their minds are made up, with 26 percent saying they are leaning in one direction and 24 percent said they have not decided.

I'm glancing though this poll and virtually everything they measure is considerably more pro-Republican than other polls.

This is part of why it's important to not care too much about what an individual polls says and to use aggregations instead.

 

[dpic73]

I'm glancing though this poll and virtually everything they measure is considerably more pro-Republican than other polls.

This is part of why it's important to not care too much about what an individual polls says and to use aggregations instead.

Harvard-Harris has consistently proven to be an outlier the last couple of years and hasn't been very accurate to boot.

The Harvard-CAPS / Harris Poll is a monthly public opinion collaboration, but its accuracy and question wording are subjects of ongoing debate among pollsters and data analysts.

• • Question Wording: Critics and election analysts often argue that the poll uses leading or complex questions designed to frame issues in a specific way, which can skew particular results compared to more traditional, neutral polling aggregates.
  • Partisan Perception: Because the poll is co-directed by prominent strategist Mark Penn (a former Democratic pollster who later became a prominent critic of modern Democrats and a corporate executive), some political analysts view its interpretations and stylistic framing as leaning conservative or sensationalized relative to academic standards.
  
  Harvard- Harris

 

[baltimorened]

I'm glancing though this poll and virtually everything they measure is considerably more pro-Republican than other polls.

This is part of why it's important to not care too much about what an individual polls says and to use aggregations instead.

you might be right, but aggregations of erroneous polls doesn't seem likely to produce an accurate one. but I understand your point

 

[baltimorened]

Harvard-Harris has consistently proven to be an outlier the last couple of years and hasn't been very accurate to boot.

The Harvard-CAPS / Harris Poll is a monthly public opinion collaboration, but its accuracy and question wording are subjects of ongoing debate among pollsters and data analysts.

• • Question Wording: Critics and election analysts often argue that the poll uses leading or complex questions designed to frame issues in a specific way, which can skew particular results compared to more traditional, neutral polling aggregates.
  • Partisan Perception: Because the poll is co-directed by prominent strategist Mark Penn (a former Democratic pollster who later became a prominent critic of modern Democrats and a corporate executive), some political analysts view its interpretations and stylistic framing as leaning conservative or sensationalized relative to academic standards.
• 
  Harvard- Harris

and, of course, the other polls are are accurate and don't skew the questions to lead to a predetermined conclusion...OK

But, as the topic of this thread indicates, I find it hard to believe that democrats are not well poised to take control of at least one chamber of the congress

 

[FLaw47]

you might be right, but aggregations of erroneous polls doesn't seem likely to produce an accurate one. but I understand your point

Can you help me understand where aggregate polling at a national level has actually been erroneous recently? Because to me it sounds like you're cherry-picking data to fit your narrative.
I do understand that some state-level polling can be off, though for actual elections I don't remember one where the aggregates have been very far off. And polling primaries is actually quite difficult.

 

[scotchtiger]

Can you help me understand where aggregate polling at a national level has actually been erroneous recently? Because to me it sounds like you're cherry-picking data to fit your narrative.
I do understand that some state-level polling can be off, though for actual elections I don't remember one where the aggregates have been very far off. And polling primaries is actually quite difficult.

Just for reference from the last even year election:

Across those six aggregators, the simple average was approximately:

Harris 48.5% — Trump 47.7% → Harris +0.8 points.

Compared with the result

The certified national popular vote ultimately came out approximately:

Trump 49.8% — Harris 48.3% → Trump +1.5 points. The Electoral College result was Trump 312, Harris 226.

So the consensus polling margin was Harris +0.8, while the actual popular-vote margin was Trump +1.5—a net miss of about 2.3 percentage points in the margin toward Harris.

 

[FLaw47]

Just for reference from the last even year election:

Across those six aggregators, the simple average was approximately:

Harris 48.5% — Trump 47.7% → Harris +0.8 points.

Compared with the result

The certified national popular vote ultimately came out approximately:

Trump 49.8% — Harris 48.3% → Trump +1.5 points. The Electoral College result was Trump 312, Harris 226.

So the consensus polling margin was Harris +0.8, while the actual popular-vote margin was Trump +1.5—a net miss of about 2.3 percentage points in the margin toward Harris.

Which is within the margin of error on those aggregates and individual polls, right?

 

[dpic73]

Just for reference from the last even year election:

Across those six aggregators, the simple average was approximately:

Harris 48.5% — Trump 47.7% → Harris +0.8 points.

Compared with the result

The certified national popular vote ultimately came out approximately:

Trump 49.8% — Harris 48.3% → Trump +1.5 points. The Electoral College result was Trump 312, Harris 226.

So the consensus polling margin was Harris +0.8, while the actual popular-vote margin was Trump +1.5—a net miss of about 2.3 percentage points in the margin toward Harris.

In polling world, this is called accurate as most polls have a margin of error of around ~ 3%.

Also, national polling isn't predictive of the Electoral College, only the popular vote.

 

[dpic73]

and, of course, the other polls are are accurate and don't skew the questions to lead to a predetermined conclusion...OK

But, as the topic of this thread indicates, I find it hard to believe that democrats are not well poised to take control of at least one chamber of the congress

Many polls have slight leans one way or the other, hence why the aggregate polling is usually more accurate. That polling has been consistent for almost a year now.

Real Clear Politics


New York Times



Silver Bulletin

 

[scotchtiger]

In polling world, this is called accurate as most polls have a margin of error of around ~ 3%.

Also, national polling isn't predictive of the Electoral College, only the popular vote.

Just using it as a reference. It’s worth noting not only the 2.3% miss, but the fact that zero aggregate polls, not a single one, correctly predicted the outcome of the popular vote.

 

[scotchtiger]

Which is within the margin of error on those aggregates and individual polls, right?

Correct on the individual polls. Not sure they list a margin of error on aggregates, but haven’t looked. I think the more interesting part is that 100% of them incorrectly predicted the outcome. You would expect some to miss high, some to miss low. But they all missed in the same direction, which has been a trend.

 

[FLaw47]

Correct on the individual polls. Not sure they list a margin of error on aggregates, but haven’t looked. I think the more interesting part is that 100% of them incorrectly predicted the outcome. You would expect some to miss high, some to miss low. But they all missed in the same direction, which has been a trend.

I don't think anybody has ever claimed that the polling is perfect, but to say that a 2.5 percentage point miss is endemic of a problem is a real stretch to me. I seem to remember the polls were actually less accurate in 2020, but nobody seemed to care because they correctly "picked the winner." Fundamentally, polls aren't trying to predict winners and losers; they're trying to predict how many people are feeling certain ways.

 

[dpic73]

Just using it as a reference. It’s worth noting not only the 2.3% miss, but the fact that zero aggregate polls, not a single one, correctly predicted the outcome of the popular vote.

Again, in polling world, this is not considered a miss as they all predicted a tight race, which it was in the popular vote. It also must be noted that pollsters have always found it difficult to project Trump's actual numbers due to shy voters and nonresponse bias (Trump voters mistrust media and pollsters).

 

[FLaw47]

Again, in polling world, this is not considered a miss as they all predicted a tight race, which it was in the popular vote. It also must be noted that pollsters have always found it difficult to project Trump's actual numbers due to shy voters and nonresponse bias (Trump voters mistrust media and pollsters).

To add to this (and I'm not going to look it up because I don't think the people I'm trying to convince will really care about the data), my recollection is that, since Trump, elections where Trump is on the ballot of had polling that somewhat skewed in Democrats favor but the opposite was true in midterms.

 

[scotchtiger]

I don't think anybody has ever claimed that the polling is perfect, but to say that a 2.5 percentage point miss is endemic of a problem is a real stretch to me. I seem to remember the polls were actually less accurate in 2020, but nobody seemed to care because they correctly "picked the winner." Fundamentally, polls aren't trying to predict winners and losers; they're trying to predict how many people are feeling certain ways.

I don’t think I made that point. My original post literally said “just for reference” with no conclusion.

I do think the trend is useful though. Here’s the 538 aggregate miss on Trump’s popular vote share the last 3 elections. Looks like they got better, but still consistently miss in the same direction.

2016: +4.3
2020: +3.4
2024: +3.0

 

[dpic73]

To add to this (and I'm not going to look it up because I don't think the people I'm trying to convince will really care about the data), my recollection is that, since Trump, elections where Trump is on the ballot of had polling that somewhat skewed in Democrats favor but the opposite was true in midterms.

This is true. It might be time to give more weight to prediction markets as they came around to Trump five months before the election.

 

[scotchtiger]

To add to this (and I'm not going to look it up because I don't think the people I'm trying to convince will really care about the data), my recollection is that, since Trump, elections where Trump is on the ballot of had polling that somewhat skewed in Democrats favor but the opposite was true in midterms.

Sort of. Still skewed to dems, but by a lower %. This is for 2022.


Thus Republicans performed roughly 1.5 points better on the D–R margin than the final 538 generic ballot suggested.

 

[scotchtiger]

Btw not saying the republicans will do well this election. I think they will get waxed.

 

[FLaw47]

Sort of. Still skewed to dems, but by a lower %. This is for 2022.


Thus Republicans performed roughly 1.5 points better on the D–R margin than the final 538 generic ballot suggested.

EDIT: I'm adding this to the top because I want to lead with my conclusion. My overall thesis was instructive and the actual data suggests to me that national polling still has plenty of value. There's just two things:

• We've had a lot of close elections recently so even accurate polling looks inaccurate when it "predicts" the wrong outcome
• There's an awful lot of laziness and data illiteracy floating around (not from you in this case, Scotch) and I hate a bad argument.

Original:
Alright, here's what I pulled from every election since 2016. For the midterms I used total house vote (imperfect though it may be). I'll provide what ChatGPT told me for 538 and RCP. The highlights are that you were basically correct. Midterms didn't really have a GOP skew (as a pattern) but they were notably more accurate.

Election	538 final average	Actual national result	Polling skew
2016 President	Clinton D+3.9	Clinton D+2.1	D+1.8
2018 House	Democrats D+8.6	Democrats D+8.6	~0.0
2020 President	Biden D+8.4	Biden D+4.5	D+3.9
2022 House	Republicans R+1.2	Republicans R+2.7	D+1.5
2024 President	Harris D+1.2	Trump R+1.5	D+2.7

Election	RCP polling margin	Actual national margin	Polling skew
2016 President	Clinton D+3.2	Clinton D+2.1	D+1.1
2018 House	Democrats D+7.3	Democrats D+8.4	R+1.1
2020 President	Biden D+7.2	Biden D+4.5	D+2.7
2022 House	Republicans R+2.5	Republicans R+2.8	D+0.3
2024 President	Harris D+0.1	Trump R+1.5	D+1.6

 

[dpic73]

Just to throw it out there, for the 2024 election, AtlasIntel was widely acclaimed as the most accurate of all pollsters.



They haven't done a Trump approval poll since May but it was also in line with the current aggregate polling.

According to data collected from May 4 to May 7, 2026:

• Disapprove: 59.8%
• Approve: 39.5%

USA: National | AtlasIntel

National-level poll on presidential approval, key topics, the economy and foreign policy. Data collection: 04-07 May 2026; Sample size: 2,069 american adults; +

www.atlasintel.org

 

[baltimorened]

Can you help me understand where aggregate polling at a national level has actually been erroneous recently? Because to me it sounds like you're cherry-picking data to fit your narrative.
I do understand that some state-level polling can be off, though for actual elections I don't remember one where the aggregates have been very far off. And polling primaries is actually quite difficult.

depending on your definition of "recently". But didn't all the polls predict a Hilary victory in 2016? 98%chance of Hilary victory. Or 2024, which polls predicted a Trump landslide?

I don't think anybody has ever claimed that the polling is perfect, but to say that a 2.5 percentage point miss is endemic of a problem is a real stretch to me. I seem to remember the polls were actually less accurate in 2020, but nobody seemed to care because they correctly "picked the winner." Fundamentally, polls aren't trying to predict winners and losers; they're trying to predict how many people are feeling certain ways.

so, if I read your post correctly, anybody pushing the results of polls on here is just blowing smoke on what could be the potential result? And politicians pay to be misled?

 

[FLaw47]

depending on your definition of "recently". But didn't all the polls predict a Hilary victory in 2016? 98%chance of Hilary victory. Or 2024, which polls predicted a Trump landslide?

Virtually everyone was predicting a Clinton victory but the polls weren't particularly inaccurate. Some media organizations used those polls to say that there was a 90% chance of Clinton winning. 538, I believe, gave Trump a 30% chance of winning. And 30% chances happen all the time. If they didn't, there'd be no point in having batters in baseball.

so, if I read your post correctly, anybody pushing the results of polls on here is just blowing smoke on what could be the potential result? And politicians pay to be misled?

I'm not sure what point you're trying to make here. I'm saying that there are a lot of people who fundamentally don't understanding polling or even basic statistics, but they still go around (much like some of the media did in 2016) proclaiming things that don't really make any sense.

 

[FLaw47]

I want to make this a little more simple for my stubborn friends. If a .300 hitter gets a base hit, were they not a .300 hitter? If Steph Curry is going to a take a free throw and I said "he's got a 90% chance of hitting this" and he misses, was I wrong?

 

[dpic73]

depending on your definition of "recently". But didn't all the polls predict a Hilary victory in 2016? 98%chance of Hilary victory. Or 2024, which polls predicted a Trump landslide?


so, if I read your post correctly, anybody pushing the results of polls on here is just blowing smoke on what could be the potential result? And politicians pay to be misled?

Reminder that national polls only predict the popular vote, not the Electoral College, so they weren't wrong since she got 3 million more votes than Trump.

 

[baltimorened]

Reminder that national polls only predict the popular vote, not the Electoral College, so they weren't wrong since she got 3 million more votes than Trump.

and all this time when the experts projected Hillary had a 98% chance of becoming president, I thought they meant she had a 98% chance of becoming Presdient. I have to admit it never occurred to me that they were predicting she had a 98% chance of winning the popular vote and not becoming president.

Just to be clear, I don't trust any of these polls....and even some of the pollster company execs don't believe the results of polls. But, hey, if you want to defend the results, I have no problems with that.

 

[dpic73]

and all this time when the experts projected Hillary had a 98% chance of becoming president, I thought they meant she had a 98% chance of becoming Presdient. I have to admit it never occurred to me that they were predicting she had a 98% chance of winning the popular vote and not becoming president.

Just to be clear, I don't trust any of these polls....and even some of the pollster company execs don't believe the results of polls. But, hey, if you want to defend the results, I have no problems with that.

The national polling was correct in 2016 but the projections that prognosticators made based on those polls missed the mark because he unexpectedly won blue wall states Michigan, Wisconsin and Pennsylvania - though their error was understandable because 2016 was only the 5th time in our history that a candidate won the popular vote but lost the Electoral College.

I alluded to it earlier but a big reason the pollsters missed was because blue collar workers are hard to reach and they made up a huge portion of his base and we've learned since then that Trump voters are notorious for being shy when speaking to pollsters because of the notion that voting for Trump was socially unacceptable, therefore many of them lied.

You can read more about it here, if interested.

Why 2016 election polls missed their mark

There is a great deal of speculation but no clear answers as to the cause of the disconnect, but there is one point of agreement: Across the board, polls underestimated Trump's level of support.

www.pewresearch.org